The Roberts Court has been restoring the separation of powers in the administrative state by strengthening presidential control over the executive branch, pressing Congress to speak clearly when it delegates major authority, and insisting that judges—not agencies—say what the law means. This judicial duty is weakened by doctrines that require courts to defer to administrative agencies interpretations of the law. When an agency and a citizen disagree about what a federal law means, judicial deference to the agency can be decisive. That deference affects not just lawyers but also citizens subject to the rules governing energy, health care, finance, and immigration.

　　One of the Roberts Court’s most important moves was ending the Chevron precedent, which often required judges to accept an agency’s “reasonable” reading of an unclear statute even when the judges thought a different reading was better. That kind of deference put the fox in charge of the henhouse, the fox being the bureaucracy and the hens being us. In plain terms, it put a thumb on the scale in favor of the government in disputes over the meaning of the law. The government had an inherent advantage over citizens because only the interpretations of government experts, not those of citizen experts, would receive deference.

　　The landmark 2024 case Loper Bright v. Raimondo overruled Chevron deference to agency interpretation of statutes and held that courts must exercise independent judgment on questions of law. Yet, almost immediately, many courts and commentators claimed that a softer form of deference to agencies ostensibly mandated by the older case, Skidmore v. Swift, survived. That view of Skidmore rests on the notion that courts may give an agency’s interpretation “weight” because of the agency’s expertise and deliberation. We can call this understanding of Skidmore “deferential Skidmore.” But if that still means extra weight for agencies, the playing field remains tilted: the government’s interpretation starts with a built-in advantage that private parties never get. Phil Pillari and I recently wrote an article showing why deferential Skidmore cannot stand, given its opposition to Loper Brights reasoning, and that a different understanding, one we can call “educational Skidmore,” is the better approach.

　　Skidmores status is the next major test of what Loper Bright means in practice: Will courts reclaim responsibility for legal interpretation, or will agencies retain an advantage? Already, lower courts are divided over whether they can apply Skidmore deference to agency interpretations. Loper Bright was unclear about the status of Skidmore deference. The Loper Bright Court does say that agency interpretations of statutes deserve “respect.” A respectful hearing is compatible with neutrality. But deference is not.

　　To be sure, a Skidmore deference regime differs from Chevron in both its rationale and its operation. Skidmore is rooted in an agency’s presumed policy expertise, while Chevron was based on the inference that any ambiguity implicitly delegates interpretive authority to the agency from Congress. And under Chevron, once an ambiguity was found, the Court was to defer absolutely to any agency interpretation so long as it was reasonable. In contrast, Skidmore deference, at least as interpreted by courts in this century, gave agency interpretations variable weight in the interpretative judgment, based on a variety of factors, such as those suggesting serious deliberation by the agency.

　　Yet, under this view, Skidmore remains a form of deference. Agency interpretations get institutional weight because the decision comes from an agency rather than because of the interpretation’s independently persuasive reasons and support. “Deferential Skidmore” confers an advantage that other litigants systematically lack, regardless of their credibility. In some cases, it may result in regulations being upheld that a completely independent review would have invalidated.

　　The argument for expertise-based deference faces a significant obstacle. Expertise can be explained. If experts within the agencies could educate those responsible for promulgating the regulation, then they could do the same with the courts. In other words, there is no reason to believe that agencies cannot fully communicate any expertise they have to evaluate, like any other expert witness. “Educational Skidmore” rather than “deferential Skidmore” would permit agencies expertise to be a source of useful illumination, not of unique authority.

　　Loper Bright shows the Court how to make the judiciary the umpire who calls balls and strikes, rather than allowing one of the players to shape the strike zone.

　　Another way to approach this point is to consider the movement of experts from government to the private sector, which often occurs in the revolving door between the regulators and the regulated. If an agency expert provides an analysis of the proper way to interpret technical terms in a statute, there is a question of why her contribution should receive more deference when she is in government service than when she leaves. The level of expertise remains unchanged.

　　But deferential Skidmore not only confuses expertise with the source of expertise, but it also conflicts with the very section of the Administrative Procedure Act (the basic statute that governs federal agencies) that Loper Bright said was inconsistent with Chevron. Section 706 provides: “the reviewing court shall decide all relevant questions of law, interpret constitutional and statutory provisions, and determine the meaning or applicability of the terms of an agency action.” The text and structure reject a standard of review for legal questions that defers to agency expertise beyond what its persuasiveness warrants. APA tells courts to decide the law, not to weigh the government’s view more heavily because it is the government’s. Congress specified deference where it wanted it (on the facts) and omitted it where it did not (on the law). Nothing in the pre-APA understanding of Skidmore supports reading § 706 as entrenching agency deference.

　　And, as we show, before the APA, Skidmore itself was not a case for general deference, but rather one focused on protecting employers peculiar reliance interests in avoiding liability for complying with wage-and-hour regulations. Skidmore gained saliency as a kind of poor man’s Chevron. As the Court made clear that Chevron did not apply (as when the agency failed to operate with the requisite formality), the government sought and increasingly received Skidmore deference.

　　Even if the APA were clearer than it is, it should be read to avoid the constitutional problems that would arise from deference to government expertise. It is anomalous and in tension with due process to give deference to only one side for the expertise they proffer. The scales of justice are not portrayed with the weight tilting toward the government’s side.

　　It might be thought, however, that even if Skidmore is wrong as a doctrine of deference, it should be protected by the doctrine of stare decisis, which applies particularly to cases decided on statutory grounds. But this view also conflicts with Loper Bright, which overruled Chevron, which, like Skidmore, was a doctrine of statutory deference.

　　If the Court found that stare decisis did not warrant preserving Chevron, the case for extending stare decisis to Skidmore is even weaker. Deferential Skidmore, like Chevron, is egregiously incorrect, unworkable, and attracts little reliance, all of which were key considerations for the Loper Bright Court in overruling Chevron.

　　Deferential Skidmore directly conflicts with the APA. At least, Chevron was grounded in the claim that Congress intended deference. Instead, Skidmore is grounded in the claim of agency expertise. If Chevron was egregiously wrong, so a fortiori is Skidmore, because that expertise, if communicated within the agency, can also be communicated to the Court and assessed on its own merits.

　　Moreover, the deferential Skidmore test fails as a workable legal test because it asks judges to assign unspecified “weight” to open-ended factors and thus invites unpredictable outcomes. Precisely because of its amorphousness, Skidmore also does not induce reliance. A doctrine whose application depends on subjective judicial assessments cannot induce reliance.

　　A further stare decisis factor not relevant to the treatment of Chevron also dooms Skidmore. It is well established that cases based on mistaken factual assumptions are more likely to be overruled when decided. But two of the justifications for deferential Skidmore have eroded over time. First, the faith that agency interpretations are rooted in expertise has dissipated as the reality of partisan political control has been broadly acknowledged. Second, judges now have far better access to the same sources agencies use, including technical materials, dictionaries, and large searchable databases of language in context.

　　The Roberts Court has made significant progress in making the administrative state conform to the Constitution, whereas its predecessors did the opposite, making the Constitution conform instead to the felt necessities of the administrative state. But with Skidmore, courts still defer to administrative agencies rather than say what the law is, acting as a neutral arbiter between the states and their citizens. Loper Bright shows the Court how to make the judiciary the umpire who calls balls and strikes, rather than allowing one of the players to shape the strike zone.